Commit
•
f202ae1
1
Parent(s):
ab7bbcf
Update README.md
Browse files
README.md
CHANGED
|
@@ -9,6 +9,15 @@ tags:
|
|
| 9 |
- merge
|
| 10 |
license: llama3
|
| 11 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
<!DOCTYPE html>
|
| 13 |
<style>
|
| 14 |
|
|
|
|
| 9 |
- merge
|
| 10 |
license: llama3
|
| 11 |
---
|
| 12 |
+
<p><h2>ExLlamaV2 Quantization</h2></p>
|
| 13 |
+
<p>Quantized with the default exllamav2 calibration dataset. Try this if you want a slightly different flavor than the RP calibrated (PIPPA) quants, with more emphasis on logic than emotion.</p>
|
| 14 |
+
|
| 15 |
+
[2.5 Bits Per Weight](https://huggingface.co/UnstableLlama/L3-MS-Astoria-70b-exl2-default-cal/tree/2_5)
|
| 16 |
+
|
| 17 |
+
[4.65 Bits Per Weight](https://huggingface.co/UnstableLlama/L3-MS-Astoria-70b-exl2-default-cal/tree/4.65)
|
| 18 |
+
|
| 19 |
+
---
|
| 20 |
+
|
| 21 |
<!DOCTYPE html>
|
| 22 |
<style>
|
| 23 |
|